In this amended application a research program is proposed for a special population of problem children: those who exhibit conduct disorders, learning delays, and problems of social interaction severe enough to jeopardize a successful placement in normal public school classrooms. Previous research in this program has developed a set of powerful procedures to be applied in special preschool classes and subsequently in the children's public school placements. These procedures have succeeded in producing acceptable classroom behavior, regular improvements in the cognitive skills taught there and appropriate peer interaction. Yet the generalization and maintenance of these critical skills has sometimes been problematic. Thus the current research proposal focuses on problems of generalization--the production, extension and maintenance of these problem childrens' behavior across settings, time, behaviors, peers, and teachers. An analysis of generalization and maintenance will be conducted within the forum of peer-mediated interventions. It is suggested that instance of generalization may represent instance of direct environmental support (such as peer reinforcement); support that is sometimes subtle and not always readily observed. This research program proposes to investigate the effectiveness of peer-mediated interventions and the potentially powerful forms of support that peers may provide in the generalization of children's behavior. Studies are proposed to determine why peer-mediated intervention work, when they do; why they effect changes in the peers who monitor as well as in the peers who are monitored; and how, when and where peers provide support for one another's behavior change. The specific content of these questions is important in its own right, as a potential contribution to clarity and analysis in the area of peer interaction and intervention. In addition, questions like these can have an important theoretical function in the study of generalization and maintenance of behavior change. They may uncover a set of direct environmental supports for what may well prove to be a directly supported process; if so the concept of generalization will no longer require more elaborate, less parsimonious, and less verifiable theoretical explanations. In sum, six studies are proposed--each clinically useful and theoretically relevant to an analysis and understanding of generalization and peer-mediation.